Characterizing the Lyman-$α$ flux probability distribution function using Legendre polynomials

The Lyman-$α$ forest is a highly non-linear field with a lot of information available in the data beyond the power spectrum. The flux probability distribution function (PDF) has been used as a successful probe of small-scale physics. In this paper we argue that measuring coefficients of the Legendre polyonomial expansion of the PDF offers several advantages over measuring the binned values as is commonly done. In particular, $n$-th coefficient can be expressed as a linear combination of the first $n$ moments, allowing these coefficients to be measured in the presence of noise and allowing a clear route for marginalisation over mean flux. Moreover, in the presence of noise, our numerical work shows that a finite number of coefficients are well measured with a very sharp transition into noise dominance. This compresses the available information into a small number of well-measured quantities. We find that the amount of recoverable information is a very non-linear function of spectral noise that strongly favors fewer quasars measured at better signal to noise.